A three-layer integrated storage method for hydrological survey multi-source data

Abstract

This invention relates to the field of hydrological informatization and hydrological data processing, and discloses a three-layer integrated storage method for multi-source hydrological survey data. The method receives multi-source hydrological survey data, generates hydrological observation event identifiers, and stores the original data in the original evidence layer using an append-only method. It standardizes and performs quality checks on the hydrological observation events, generating intermediate version records stored in the intermediate version layer. Based on the valid intermediate version records, it generates application data partitions and organizes them into application snapshots. After identifying late or revised events, it determines the target failure set by combining the corresponding river network topology version, hydrological propagation time, and data lineage credentials. Incremental recalculation is initiated from the valid model state checkpoint, and the dynamic recalculation closure boundary is determined based on the model state convergence result. The recalculation result is written to a shadow application snapshot, which is then switched to a valid application snapshot after consistency verification.

Description

Technical Field

[0001] This invention relates to the field of hydrological informatization and hydrological data processing, specifically to a three-layer integrated storage method for multi-source hydrological survey data. Background Technology

[0002] With the continuous development of hydrological monitoring equipment, remote sensing platforms, mobile flow measurement equipment, and hydrological operational systems, the sources and forms of data generated during hydrological surveys are constantly increasing. This data includes both structured time-series data such as rainfall, water level, flow rate, and velocity, and unstructured data such as raw reports, measurement records, images, videos, equipment logs, and historical compiled documents. To meet the operational needs of rainfall and water level monitoring, flood forecasting, water resources analysis, and statistical report generation, it is typically necessary to receive, store, quality-check, perform model calculations, and publish these data for application.

[0003] Existing hydrological data management systems typically set up separate databases for raw data, results, and operational applications. Raw data is written to the results database after format conversion, quality control, or manual review, and then read by various operational systems. These systems often use current valid values ​​as the basis for storage and calculation. When automatic station data is delayed, manual measurement results are supplemented, equipment calibration results change, or historical data is revised after review, the revisions are often saved by updating existing records. This makes it difficult to completely preserve the version relationships between raw data, quality control results, and application results. At the same time, structured measurements are often stored separately from unstructured data such as raw reports and test documents. There is a lack of clear correlation between application results and the actual data versions they reference, easily leading to the problem of retrieving data from different processing stages in the same operational query.

[0004] For late or revised data, existing systems typically recalculate results for relevant time periods, stations, or the entire watershed based on dependencies between database tables. Relying solely on data dependencies fails to reflect the actual spatial direction and temporal range of hydrological data propagation along the river network, easily including data without actual hydrological impact in the recalculation. Conversely, relying solely on upstream and downstream river network relationships makes it difficult to determine whether the applied results actually referenced the revised data. Especially when hydrological models exhibit state continuity characteristics, existing methods struggle to determine a reasonable recalculation starting point and when the revision's impact has disappeared, often requiring large-scale recalculations starting from earlier points, increasing the burden on data reading, cross-database transfer, and model operation. Therefore, there is an urgent need for a multi-source hydrological survey data storage method that can maintain multi-layered data version relationships and control the recalculation scope of late or revised data based on the hydrological propagation range and actual data reference relationships. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a three-layer integrated storage method for multi-source hydrological survey data, thereby resolving the technical problems existing in the prior art.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: A three-layer integrated storage method for multi-source hydrological survey data includes the following steps: S1: Receive multi-source hydrological survey data, parse hydrological spatial object identifiers, observation element identifiers, observation times, data source identifiers, and source data versions, generate hydrological observation event identifiers, and bind the hydrological observation event identifiers to the river network topology versions corresponding to the observation times; S2: Store the original data corresponding to the hydrological observation event in the original evidence layer in an append-only manner, perform standardization processing and quality verification on the hydrological observation event, and generate intermediate version records stored in the intermediate version layer; S3: Based on intermediate version records whose quality status is provisionally valid or officially valid and not marked as invalid, generate application data partitions according to hydrological nodes and calculation periods, organize the application data partitions into application snapshots stored in the application snapshot layer, and generate data lineage certificates that record the intermediate version records they reference and model state checkpoints that record the corresponding set of hydrological node model state variables for each application data partition. S4: When a late event or revision event is identified, a new intermediate version record is generated, the changed intermediate version record is identified as being supplemented or replaced by the new intermediate version record, and the hydrological propagation cone composed of the affected hydrological nodes and their time intervals is determined based on the river network topology version, river direction and hydrological propagation time corresponding to the late event or revision event. S5: The application data partition whose hydrological node and calculation period are located within the hydrological propagation cone and whose data lineage certificate directly or indirectly references the changed intermediate version record is identified as the target failure set; S6: Start incremental recalculation from the nearest model state checkpoint that is before the start time of the impact of the target failure set, does not reference the intermediate version record affected by the change, and uses the same hydrological model version and river network topology version as this incremental recalculation. Determine the dynamic recalculation closure boundary based on whether the new model state variable set formed by the incremental recalculation and the old model state variable set in the original application snapshot continuously meet the convergence condition, and stop recalculating the application data partition after the dynamic recalculation closure boundary. S7: Write the application data partitions formed by incremental recalculation into the shadow application snapshot, and reference the application data partitions in the original application snapshot that do not belong to the target failure set to the shadow application snapshot. After the shadow application snapshot passes the closed boundary state and version relationship consistency verification, switch the effective application snapshot from the original application snapshot to the shadow application snapshot.

[0007] Preferably, multi-source hydrological survey data comes from automatic hydrological stations, manual measurement terminals, mobile flow measurement equipment, remote sensing data platforms, and historical compiled data; The hydrological spatial object identifier, observation element identifier, observation time, data source identifier, and source data version are concatenated according to a fixed field order. The concatenation result is then used to perform a summary operation on the data content summary of the multi-source hydrological survey data to obtain the hydrological observation event identifier.

[0008] Preferably, when storing the original data in the original evidence layer, the structured measurements are written into the time series data area according to the hydrological spatial object identifier, the observation element identifier and the observation time, the unstructured data is written into the object storage area, and an event index is established with the hydrological observation event identifier as the index key. The event index records the data partition where the structured measurement value is located, the object address and data content summary of the unstructured data, the data reception time, and the data entry batch; For new received data with the same hydrological spatial object identifier, observation element identifier, and observation time, the new hydrological observation event is written into the original evidence layer without covering the existing hydrological observation event.

[0009] Preferably, the standardization process includes unit conversion, time base conversion, spatial object encoding mapping, and missing test identifier conversion; Based on the data source category, observation element category, hydrological spatial object category, and acquisition method of hydrological observation events, quality verification items are determined from the applicable relationships of quality verification. Value range verification and time continuity verification are performed on structured measurements, change rate verification is performed on structured measurements with continuous sampling time sequence, upstream and downstream hydrological relationship verification is performed on structured measurements located at river network connectivity nodes, and equipment operation status verification is performed on structured measurements automatically acquired by monitoring equipment. If the currently executable quality verification project passes but the verification time window has not yet ended, an intermediate version record with a provisionally valid quality status is generated; when the verification time window ends and all applicable quality verification projects pass, an intermediate version record with a formally valid quality status is generated. The intermediate version record also records the verification rule version, equipment calibration version, river network topology version, and the identifier of the previous intermediate version, and marks the previous intermediate version record that is replaced by the new intermediate version record as invalid.

[0010] Preferably, for each combination of data source identifier, hydrological spatial object identifier, and observation element identifier, the time boundary of processed events and the allowable delay duration are maintained. The time boundary of processed events is the maximum observation time of the intermediate version record in the corresponding combination when the quality status reaches the provisional or formal validity. The allowable delay duration is determined according to the acquisition cycle, reporting cycle, and retransmission period of the corresponding acquisition source. When the observation time of newly received hydrological survey data is earlier than the difference between the time boundary of the processed event and the allowable lateness, it is identified as a late event, and the missing intermediate version record that has the same hydrological spatial object identifier, observation element identifier and observation time as the late event is identified as the intermediate version record to be changed. When newly received hydrological survey data and existing hydrological observation events have the same hydrological spatial object identifier, observation element identifier, and observation time, and the source data versions or data content summaries of the two are different, the newly received hydrological survey data will be identified as a revision event, and the intermediate version record corresponding to the existing hydrological observation event will be identified as the intermediate version record subject to change.

[0011] Preferably, the river network topology version records hydrological nodes, river segments, the connection relationship between nodes and river segments, river direction, river segment propagation time, and the effective time interval of the topology version; Starting from the hydrological node where the late event or revised event is located, traverse the downstream hydrological nodes along the river direction, and determine the sum of the observation time of the late event or revised event and the cumulative river segment propagation time to the current downstream hydrological node as the start time of the impact of the current downstream hydrological node. The impact termination time of the current downstream hydrological node is determined based on the basin confluence response duration, model state memory duration, cumulative application data time window length, and storage response duration of storage nodes in the propagation path. The impact time interval of the current downstream hydrological node is formed by the impact start time and impact termination time.

[0012] Preferably, the data lineage certificate also records the upstream application data partition identifier, hydrological model version, business rule version, river network topology version, and output result summary referenced by the application data partition; When a data lineage certificate record is modified by an intermediate version record, the corresponding application data partition is determined to directly reference the modified intermediate version record. When the data lineage certificate records the upstream application data partition, and the data lineage certificate of the upstream application data partition is traced step by step to obtain the changed intermediate version record, it is determined that the corresponding application data partition indirectly references the changed intermediate version record. The application data partition that directly or indirectly references the changed intermediate version record and belongs to the affected hydrological nodes in the hydrological propagation cone, is added to the target failure set.

[0013] Preferably, the model status checkpoint also records the upstream contribution data summary, hydrological model version, river network topology version, and application snapshot version; The upstream contribution data summary is generated based on the intermediate version identifier referenced by the application data partition, the upstream application data partition identifier, the output result summary of the upstream application data partition, the hydrological model version, and the river network topology version; If there are no unreferenced intermediate version records before the start time of the impact and the model status checkpoints that use the same hydrological model version and river network topology version as this incremental recalculation, the starting state of the hydrological calculation cycle containing the start time of the impact shall be used as the starting point of the incremental recalculation. Incremental recalculation is performed according to the order of calculation time periods and river direction.

[0014] Preferably, the convergence threshold of each model state variable is determined according to the allowable state deviation specified in the corresponding hydrological model version and the allowable deviation specified in the business rule version; The number of consecutive convergence calculation periods is determined based on the minimum stable duration of the hydrological model version and the model calculation step size, and the number of consecutive convergence calculation periods is not less than two. When the difference between the old and new values ​​of each model state variable is not greater than its corresponding convergence threshold, the new and old upstream contribution data summaries are consistent, and there are no unprocessed late events or revision events in each calculation period and its upstream propagation path, the last calculation period is determined as the dynamic recalculation closure boundary. When the river network has branches, the dynamic recalculation closure boundary is determined for each branch, and the incremental recalculation range is stopped from being expanded for branches that have formed dynamic recalculation closure boundaries.

[0015] Preferably, the shadow application snapshot records the application data partitions formed by incremental recalculation and the application data partitions referenced from the original application snapshot through the snapshot list; The shadow application snapshots are subjected to upstream and downstream continuity verification, water balance verification, closed boundary state verification, and version consistency verification. Among them, the upstream and downstream continuity verification determines whether the difference between the output flow of the upstream hydrological node after the river segment propagation time alignment and the input flow component of the downstream hydrological node originating from the upstream hydrological node exceeds the continuity error limit. The water balance verification determines whether the water balance residual formed by the inflow, interval inflow, outflow, water storage change and water loss exceeds the water balance error limit. The closed boundary state verification determines whether the difference between the old and new values ​​of the model state variables at the dynamically recalculated closed boundary exceeds the corresponding convergence threshold. The version consistency verification determines whether the intermediate version, hydrological model version and river network topology version recorded in the snapshot list are consistent with the data lineage certificate of each application data partition. When the closed boundary state verification fails, the incremental recalculation range is extended along the calculation period direction. When the upstream and downstream continuity verification or water balance verification fails, the application data partitions of the hydrological nodes at both ends of the river section that failed the verification are added to the target failure set within the corresponding impact time interval. When the version consistency verification fails, the snapshot list and data lineage certificate are regenerated. After all checks pass, the valid snapshot pointer is switched atomically from the original application snapshot to the shadow application snapshot. In summary, the present invention has the following main beneficial effects: This application constructs a three-layer integrated storage system consisting of an original evidence layer, an intermediate version layer, and an application snapshot layer. This system stores multi-source hydrological survey data in append-only manner based on hydrological observation events. The intermediate version records retain the quality status, processing version, and the replacement relationship between previous and subsequent data. This avoids the overwriting of original records by supplementary data, verification changes, or rule updates, enabling continuous traceability between original data, intermediate processing results, and application results, and reducing data inconsistencies caused by the mixed use of different data versions.

[0016] By combining the historical river network topology version, hydrological propagation time, and application data lineage credentials corresponding to late or revised events, a hydrological propagation cone composed of affected hydrological nodes and their time intervals is determined. Only application data partitions located within the hydrological propagation cone that directly or indirectly reference the changed intermediate version records are included in the target failure set. This achieves the effect of narrowing the failure scope according to both the hydrological physical propagation relationship and the actual data reference relationship, avoiding the duplication of irrelevant data due to relying solely on the upstream and downstream relationships of the river network, and also avoiding the expansion of recalculations that do not conform to the actual hydrological impact due to relying solely on database lineage relationships.

[0017] By initiating incremental recalculation from the model state checkpoint that is most recent and consistent with the version before the start of the impact, and determining the dynamic recalculation closure boundary based on the convergence of the state variables of the old and new models during the continuous calculation period, the effect of terminating subsequent recalculation in a timely manner after the impact of late or revised data disappears is achieved. At the same time, by writing the recalculation results to the shadow application snapshot, and switching to the effective application snapshot only after the closure boundary state, upstream and downstream continuity, water balance, and version relationship verification are passed, the effects of reducing database reads, cross-layer data transmission, and redundant calculations of hydrological models are achieved, while maintaining the integrity and version consistency of application data release. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 refer to Figure 1 A three-layer integrated storage method for multi-source hydrological survey data includes the following steps: S1: Receive multi-source hydrological survey data, parse hydrological spatial object identifiers, observation element identifiers, observation times, data source identifiers, and source data versions, generate hydrological observation event identifiers, and bind the hydrological observation event identifiers to the river network topology versions corresponding to the observation times; S2: Store the original data corresponding to the hydrological observation event in the original evidence layer in an append-only manner, perform standardization processing and quality verification on the hydrological observation event, and generate intermediate version records stored in the intermediate version layer; S3: Based on intermediate version records whose quality status is provisionally valid or officially valid and not marked as invalid, generate application data partitions according to hydrological nodes and calculation periods, organize the application data partitions into application snapshots stored in the application snapshot layer, and generate data lineage certificates that record the intermediate version records they reference and model state checkpoints that record the corresponding set of hydrological node model state variables for each application data partition. S4: When a late event or revision event is identified, a new intermediate version record is generated, the changed intermediate version record is identified as being supplemented or replaced by the new intermediate version record, and the hydrological propagation cone composed of the affected hydrological nodes and their time intervals is determined based on the river network topology version, river direction and hydrological propagation time corresponding to the late event or revision event. S5: The application data partition whose hydrological node and calculation period are located within the hydrological propagation cone and whose data lineage certificate directly or indirectly references the changed intermediate version record is identified as the target failure set; S6: Start incremental recalculation from the nearest model state checkpoint that is before the start time of the impact of the target failure set, does not reference the intermediate version record affected by the change, and uses the same hydrological model version and river network topology version as this incremental recalculation. Determine the dynamic recalculation closure boundary based on whether the new model state variable set formed by the incremental recalculation and the old model state variable set in the original application snapshot continuously meet the convergence condition, and stop recalculating the application data partition after the dynamic recalculation closure boundary. S7: Write the application data partitions formed by incremental recalculation into the shadow application snapshot, and reference the application data partitions in the original application snapshot that do not belong to the target failure set to the shadow application snapshot. After the shadow application snapshot passes the closed boundary state and version relationship consistency verification, switch the effective application snapshot from the original application snapshot to the shadow application snapshot.

[0021] In this application, the original evidence layer, intermediate version layer, and application snapshot layer are each provided with independent storage space and write permissions. The original evidence layer only allows the data access program to perform append writes; the intermediate version layer reads the hydrological observation events in the original evidence layer and writes the standardized results and quality verification results; the application snapshot layer reads the valid intermediate version records in the intermediate version layer and writes the hydrological model calculation results or business calculation results. The application snapshot layer may not modify the intermediate version layer in reverse, and the intermediate version layer may not modify or delete the original data in the original evidence layer.

[0022] Before the system goes into operation, establish the following tables: hydrological spatial object table, observation element table, data source configuration table, quality verification applicable relationship table, river network topology version table, equipment calibration version table, hydrological model version table, and business rule version table.

[0023] The hydrological spatial object table records the identifiers of hydrological stations, rain gauges, flow measurement sections, river sections, reservoirs, dams, and catchment units, and records the correspondence between hydrological spatial objects and hydrological nodes. The observation element table records the observation element identifiers, standard units, data precision, and physical value boundaries for rainfall, water level, flow rate, flow velocity, and water storage. The data source configuration table records the data source identifier, acquisition method, acquisition cycle, reporting cycle, allowable lateness, and configuration version.

[0024] The aforementioned basic data are derived from actual station archives, equipment communication protocols, equipment technical data, hydrological data compilation rules, business reporting system, hydrological model configuration files, and model calibration or acceptance data. Fixed values ​​that are detached from actual business operations are not uniformly set.

[0025] The system receives multi-source hydrological survey data from at least two types of acquisition sources, including automatic hydrological stations, manual measurement terminals, mobile flow measurement equipment, remote sensing data platforms, and historical compiled data. It also parses hydrological spatial object identifiers, observation element identifiers, observation times, data source identifiers, and source data versions from the received data.

[0026] The observation time for automatic hydrological station data is the sampling time recorded by the acquisition terminal; the observation time for manual measurement data is the actual measurement time or the start and end period of the measurement; the observation time for remote sensing data is the imaging time or observation period recorded by the remote sensing product; and the observation time for historical compiled data is the hydrological time recorded in the data. The server reception time is stored separately and is not used to replace the observation time.

[0027] The source data version should preferentially adopt the message sequence number, file version number, or result revision number provided by the acquisition source. If the acquisition source does not provide a version number, incremental source data versions should be generated according to the order of reception under the same hydrological spatial object identifier, observation element identifier, and observation time.

[0028] The system calculates a data content summary for the original data without modification, then concatenates the hydrological spatial object identifier, observation element identifier, observation time, data source identifier, source data version, and data content summary according to a fixed field order. A summary operation is then performed on the concatenation result to generate hydrological observation event identifiers. The summary operation uses a system-wide configured algorithm, and the character encoding, time format, numerical precision, and null value representation are standardized during field concatenation to ensure that the same input data generates the same hydrological observation event identifiers.

[0029] The system queries the river network topology version table based on the observation time of the hydrological survey data, identifies the river network topology version whose valid time interval includes the observation time, and establishes a binding relationship between the hydrological observation event identifier and the river network topology version. When the river network structure changes subsequently, the hydrological observation events that have already occurred will still maintain their binding with the original river network topology version corresponding to their observation time.

[0030] The raw data corresponding to hydrological observation events are written to the raw evidence layer in an append-only manner. Append-only means that the hydrological observation events that have already been written will not be overwritten or deleted due to subsequent supplementary reports, verification and modification of values, changes in equipment calibration, or changes in business rules.

[0031] Structured measurements are written into the time-series data area according to the hydrological spatial object identifier, observation element identifier, and observation time, and the hydrological observation event identifier, data source identifier, source data version, observation time, and reception time are recorded.

[0032] Raw messages, images, videos, test records, historical documents, and device logs are written as unstructured data to the object storage area. The object storage area generates an object address and an object content summary for each piece of unstructured data.

[0033] The system establishes an event index using hydrological observation event identifiers as index keys. The event index records the data partition where structured measurements are located, the object address and summary of unstructured data, the data reception time, and the data batch. When the same hydrological observation event includes both structured measurements and unstructured attachments, a correspondence between the two is established using the same hydrological observation event identifier.

[0034] After unstructured data is written, the object content digest is recalculated. If the recalculated object content digest is inconsistent with the data content digest received, the object is determined to be incomplete, the object address registration is revoked, and the data is rewritten. Incompletely written data is not passed to the intermediate version layer.

[0035] The intermediate version layer reads hydrological observation events from the original evidence layer and performs standardization processing on them. Standardization processing includes unit conversion, time reference conversion, spatial object encoding mapping, and missing data identifier conversion.

[0036] Unit conversion is performed based on the standard units and legal conversion relationships recorded in the observation element table; time reference conversion converts the observation time of different data sources into a unified system time reference while retaining the original observation time; spatial object coding mapping maps the station code or equipment code used by the acquisition source to hydrological spatial object identifiers; missing measurement identifier conversion converts the null values, special characters or missing measurement codes used by different acquisition sources into a unified missing measurement status, and the missing measurement status is not used as the actual measured value input into the hydrological model.

[0037] The system queries the applicable relationship table for quality verification based on the data source category, observation element category, hydrological spatial object category, and acquisition method of the hydrological observation event to determine the quality verification items that need to be performed for the hydrological observation event.

[0038] For structured measurements, the range and time continuity are checked; for structured measurements with continuous sampling time sequences, the rate of change between adjacent calculation periods is checked; for water level, flow, or rainfall-runoff data located at river network connectivity nodes, the upstream-downstream hydrological relationship is checked; for data automatically collected by monitoring equipment, the equipment operating status is checked. For unstructured data, file integrity, format, object content summary, and hydrological observation event correspondence are checked.

[0039] In quality verification, boundaries are divided into hard boundaries and suspected anomaly boundaries. Equipment range, the physical boundaries of observed elements, and the absolute limits stipulated in current operational rules are considered hard boundaries; data exceeding these hard boundaries are marked as failing the verification. Historical valid data distribution, historical change rates, and relationships with adjacent stations are used to determine suspected anomaly boundaries; data exceeding these suspected anomaly boundaries are marked as requiring further verification and are not directly considered invalid.

[0040] The value range, change rate limit, and upstream / downstream relationship verification conditions are all recorded in the verification rule version. When using historical data to determine the change rate limit, the start and end times of the historical data, the invalid data types that were removed, the statistical method used, the resulting limit, and the review time are recorded. When the verification rules change, a new verification rule version is generated, and the existing intermediate version records continue to retain the original verification rule version actually used.

[0041] For periodically acquired data, the system generates a sequence of expected observation times based on the acquisition period and task activation time in the data source configuration. If a certain expected observation time has passed the time boundary of the processed event and the original data has not yet been received, an intermediate version record with a quality status of "missing data pending" is generated.

[0042] For task-triggered data such as manual surveys and remote sensing products, a list of data to be received is generated based on the task identifier, task start and end times, required observation elements, and latest reporting deadline. If data on the list has not arrived by the latest reporting deadline, an intermediate version record of missing data to be supplemented is generated. For data without a fixed collection cycle and no task reporting requirements, a record of missing data to be supplemented is not automatically generated.

[0043] If all currently feasible quality checks for a hydrological observation event have been completed, but relevant verification data has not yet arrived or the verification window has not yet closed, an intermediate version record with a provisionally valid quality status is generated. When the data requiring cross-verification has arrived, the verification window has closed, and all applicable quality verification items have passed, an intermediate version record with a formally valid quality status is generated.

[0044] Each intermediate version record must include at least the intermediate version identifier, hydrological observation event identifier, standardized data, quality status, verification rule version, equipment calibration version, river network topology version, and the identifier of the previous intermediate version. Once a new intermediate version record is generated, the previous intermediate version record it replaces is marked as invalid, but it is not physically deleted.

[0045] The system generates application data based on intermediate version records whose quality status is provisionally valid or officially valid and which have not been marked as invalid. Provisionally valid data is used for real-time rainfall and water level monitoring or operations with high timeliness requirements, while officially valid data is used for hydrological compilation, statistical reports, and operations requiring stable versions. The quality status that each operation can use is determined by the version of the operation rules.

[0046] Application data is divided into application data partitions according to hydrological nodes and calculation periods. The calculation period adopts the time step actually used by the corresponding hydrological model or business calculation, and the same calculation period is used for the same hydrological model in the same application snapshot.

[0047] Each application data partition generates a data lineage certificate. The data lineage certificate records the intermediate version identifier directly referenced by the application data partition, the upstream application data partition identifier referenced, the input data time range, the hydrological model version, the business rule version, the river network topology version, the river segment identifier, and the output result summary.

[0048] When an application data partition directly uses an intermediate version record as input, it constitutes a direct reference to that intermediate version record. When an application data partition references an upstream application data partition, and that upstream application data partition directly or indirectly references an intermediate version record, it constitutes an indirect reference to that intermediate version record. Each data lineage certificate forms a directed reference chain from the input data partition to the output data partition through the upstream application data partition identifier.

[0049] Multiple application data partitions are organized into application snapshots using a snapshot list. The snapshot list records the snapshot version, data expiration time, storage location of the application data partition, intermediate version range, hydrological model version, business rule version, and river network topology version. A valid application snapshot is bound when a business query is initiated, and the bound application snapshot remains unchanged until the query ends.

[0050] Each application data partition corresponds to a model status checkpoint. The model status checkpoint records the set of model status variables, upstream contribution data summary, hydrological model version, river network topology version, and application snapshot version of the hydrological nodes at the end of the calculation period.

[0051] The set of model state variables is determined by the actual hydrological model version used, selecting the state variables that need to be inherited when running in the next calculation period. The model state variables for rainfall-runoff models may include soil moisture content, free water storage, and runoff status; the model state variables for river runoff models may include river storage, inflow status, and outflow status; and the model state variables for hydrodynamic models may include cross-sectional water level, flow rate, and velocity status. The specific state variables, units, numerical precision, and output interfaces used are recorded in the hydrological model version.

[0052] The river network topology version assigns a segment identifier to each river segment connecting upstream and downstream hydrological nodes. When the hydrological model calculates the input to downstream nodes, it saves the input flow components after propagation through the corresponding river segment to each upstream hydrological node according to the segment identifier, and then combines each input flow component, the inflow of water in the interval, and other input quantities to form the total input of downstream nodes.

[0053] The upstream contribution data summary is generated jointly by the intermediate version identifier referenced by the current application data partition, the upstream application data partition identifier, the output result summary of the upstream application data partition, the hydrological model version, and the river network topology version. Therefore, when the old and new upstream contribution data summaries are consistent, it indicates not only that the referenced objects are consistent, but also that the upstream output content and processing version are consistent.

[0054] The system maintains processed event time boundaries for each combination of data source identifier, hydrological spatial object identifier, and observation element identifier. The processed event time boundary is the maximum observation time in the intermediate version record corresponding to that combination when the quality status reaches provisional or formal validity.

[0055] The allowable delay duration is stored in the data source configuration version. The allowable delay duration for automatically monitored data is determined based on the normal upload cycle and maximum retransmission window specified in the communication protocol; for manually tested data, it is determined based on the latest reporting deadline stipulated in the business procedures; and for remote sensing products, it is determined based on the product production cycle and delivery time limit. When multiple normal arrival deadlines exist for the same data source, the latest allowed normal arrival deadline is adopted. After the data source configuration version is updated, historical data still uses the original data source configuration version corresponding to its observation time.

[0056] When the observation time of newly received data is earlier than the difference between the time boundary of the processed event and the allowable delay time, the corresponding hydrological observation event is identified as a late event. The system queries for missing intermediate version records that have the same hydrological spatial object identifier, observation element identifier, and observation time as the late event, and identifies the intermediate version record that has been modified.

[0057] When newly received data and existing hydrological observation events have the same hydrological spatial object identifier, observation element identifier, and observation time, but the source data versions or data content summaries are different, the hydrological observation event corresponding to the newly received data is identified as a revision event, and the intermediate version record corresponding to the existing hydrological observation event is identified as the modified intermediate version record.

[0058] When the source data version and data content summary of newly received data are consistent with existing hydrological observation events, it is considered as duplicate data reception. Only duplicate reception logs are recorded, no new intermediate version records are generated, and incremental recalculation is not initiated.

[0059] After standardization and quality verification, late or revised events are recorded as new intermediate versions, and a supplementary or alternative relationship is established between the new intermediate version record and the changed intermediate version record.

[0060] The river network topology version records hydrological nodes, river segments, the connections between nodes and river segments, river direction, river segment propagation time, water storage nodes, river segment identifiers, and the effective time interval of the topology version. When a station is relocated, a cross-section is adjusted, a river segment is re-divided, or the water storage relationship changes, a new river network topology version is generated, while the original river network topology version is retained.

[0061] The propagation time of a river segment should be based on the measured peak propagation time or the tracing process propagation time, which has been verified through operational procedures. When measured data is unavailable, the calculation results of a calibrated and validated hydrological or hydrodynamic model should be used. When the same river segment has propagation parameters corresponding to different flow levels, the corresponding parameters should be selected based on the flow level at the time of the delayed or revised event; when interpolation is required, the interpolation method specified in the river segment propagation parameter table should be used.

[0062] The river section propagation time only includes the water flow propagation time within the river section, and does not include the residence time caused by reservoir or dam regulation. When the river section propagation parameters actually used already include the regulation effect, the corresponding regulation response time will not be added separately.

[0063] The storage and regulation response duration is determined based on the reservoir water level, gate status, and the version of the dispatching rules at the time of the event. The basin runoff response duration is obtained from audited unit hydrograph data, historical rainfall-runoff processes, or calibrated rainfall-runoff models. The cumulative time window length for application data adopts the statistical period specified in the business rules version.

[0064] The model state memory duration is recorded in the hydrological model version. When determining the model state memory duration, historical hydrological processes covering low, medium, and high water levels are selected from the model calibration or validation period. A perturbation of the same magnitude as the actual revision is applied to the input data or initial state of the historical process, and the model is run before and after the perturbation. When the difference between the model state variables falls within the operationally permissible error range and does not exceed this range in subsequent calculations, the time elapsed from the occurrence of the perturbation to state stabilization is determined as the state memory duration of that historical process, and the maximum value in the selected historical process is taken as the model state memory duration adopted for this hydrological model version.

[0065] After a late event or a revised event forms a new intermediate version record, the system calls the river network topology version bound to the event's observation time, starting from the hydrological node where the event is located and traversing downstream hydrological nodes along the river direction.

[0066] For any affected hydrological node, the start time of its impact is determined based on the event observation time and the cumulative river segment propagation time from the hydrological node where the event occurred to the affected hydrological node: ; In the formula, Indicates the affected hydrological nodes The impact begins at the time of the event; Indicates the observation time for late or revised events; This indicates the distance from the hydrological node where the event occurred to the affected hydrological nodes. The propagation path; Indicates the river segment in the propagation path; Indicates river section The spread time.

[0067] The impact termination time is determined based on the impact start time, the basin confluence response duration, the model state memory duration, the cumulative application data time window length, and the storage response duration in the propagation path. The confluence response duration, model state memory duration, and cumulative application time window continue in parallel after the impact reaches the node; therefore, the longest duration is used. When the propagation path passes through different storage nodes in sequence, the additional delay generated by each storage node is included separately.

[0068] When a river network has branches or multiple propagation paths, the earliest start time of influence among all effective propagation paths is taken as the start time of influence for that hydrological node, and the latest end time of influence is taken as the end time of influence for that hydrological node. All affected hydrological nodes and their corresponding influence time intervals together form a hydrological propagation cone.

[0069] The system performs space checks, time checks, and version reference checks respectively.

[0070] Spatial determination is used to determine whether the hydrological node to which the application data partition belongs is an affected hydrological node in the hydrological propagation cone. Temporal determination is used to determine whether the calculation period of the application data partition intersects with the influence time interval of the hydrological node.

[0071] When an application data partition's data lineage certificate directly records the modified intermediate version record, it is determined that the application data partition directly references the modified intermediate version record. When an application data partition references an upstream application data partition, and tracing the data lineage certificate of the upstream application data partition up to the modified intermediate version record leads to it, it is determined that the application data partition indirectly references the modified intermediate version record.

[0072] Only application data partitions that simultaneously meet the spatial, temporal, and direct or indirect reference criteria are added to the target failure set. Application data partitions located downstream of the event but whose calculation period has not yet entered the impact time interval are not added to the target failure set; application data partitions with data lineage paths but which have exceeded the hydrological propagation cone's impact termination time are also not added to the target failure set.

[0073] For each affected hydrological node in the target failure set, the system searches for model state checkpoints in reverse chronological order before the start time of the impact.

[0074] The model status checkpoint that serves as the starting point for incremental recalculation should simultaneously meet the following requirements: no reference to the modified intermediate version record; the hydrological model version used is consistent with the hydrological model version used in this incremental recalculation; and the river network topology version used is consistent with the river network topology version used in this incremental recalculation.

[0075] If no suitable model state checkpoint exists, the starting state of the hydrological calculation cycle, including the start time of the influence, is used as the starting point for incremental recalculation. The hydrological calculation cycle adopts the flood process, hydrological day, statistical cycle, or continuous model operation cycle determined in actual operations. If the hydrological model requires preheating, recalculation begins from the preheating starting point specified in that calculation cycle.

[0076] Incremental recalculation is performed in the order of calculation time periods, and the calculation output is transmitted to downstream hydrological nodes according to the river direction specified by the river network topology.

[0077] After each calculation period of a hydrological node is completed, the system generates a new set of model state variables, new application results, and a new upstream contribution data summary, and compares them with the old set of model state variables and the old upstream contribution data summary in the original application snapshot.

[0078] For each model state variable, the convergence threshold is the smaller of the allowable state deviation specified in the hydrological model version and the allowable deviation specified in the business rule version. The comparison of model state values ​​uses the uncranked state values ​​that meet the accuracy requirements of this convergence threshold, and does not rely on the rounded values ​​displayed on the interface to determine whether convergence has occurred.

[0079] The number of consecutive convergence periods is determined based on the minimum stable duration and calculation step size of the hydrological model, and must be no less than two calculation periods. The minimum stable duration is determined through the model's historical validation process. That is, after the model's state variables first meet the convergence threshold, the duration during which they remain convergent and do not exceed the threshold again is recorded, and the maximum stable duration in different validation processes is taken as the stability judgment criterion for the corresponding hydrological model version.

[0080] When the difference between the old and new values ​​of each model state variable is not greater than its corresponding convergence threshold within a specified number of consecutive calculation periods, the summaries of the old and new upstream contribution data are consistent, and there are no unprocessed late events or revision events in these calculation periods and their upstream propagation paths, the last calculation period is determined as the dynamic recalculation closure boundary.

[0081] Once the dynamic recalculation closure boundary is determined, the incremental recalculation range is stopped from expanding along the calculation time period, and the output after the closure boundary is stopped from being propagated to downstream hydrological nodes that no longer receive affected inputs. When the river network has branches, each branch determines its own dynamic recalculation closure boundary. Branches that have already closed stop expanding, while branches that have not yet closed continue to perform incremental recalculation.

[0082] The newly generated application data partitions from the incremental recalculation are written to the shadow application snapshot. The original application data partitions in the target failure set are replaced by the new application data partitions. Application data partitions in the original application snapshot that were not included in the target failure set are not recalculated, but instead a read-only reference to the application data partition is created in the snapshot list of the shadow application snapshot.

[0083] The snapshot list for shadow application snapshots records the snapshot version, data expiration time, storage location or reference location of each application data partition, intermediate version identifier, hydrological model version, business rule version, river network topology version, and data lineage certificate. Shadow application snapshots are not used as valid snapshots for business queries until consistency verification passes.

[0084] After the shadow application snapshot is formed, upstream and downstream continuity verification, water balance verification, closed boundary state verification, and version consistency verification are performed.

[0085] The upstream-downstream continuity check extracts the output flow from upstream hydrological nodes and the input components originating from upstream hydrological nodes from the input flow of downstream hydrological nodes based on the river segment identifier, and aligns them according to the river segment propagation time. When a downstream hydrological node has both tributary inflows and inter-regional inflows, the total downstream input flow is not directly compared with the output flow of a single upstream node.

[0086] The continuity error limit shall be based on the allowable error specified in the hydrological model acceptance document or quality control rules for the corresponding river section. If no such limit is specified, it shall be determined based on the uncertainty of the upstream output flow, downstream input flow components and the uncertainty of the river section propagation model, and the source of uncertainty and determination process shall be recorded in the version of the verification rules.

[0087] Water balance verification takes the input water volume, inflow water volume, output water volume, water storage changes, and water loss actually included in the hydrological model within the same calculation period as its objects. The water balance residual is: ; In the formula, This represents the water balance residual; Indicates the amount of water input from upstream; Indicates the water inflow within the interval; Indicates the output water volume; Indicates the change in water storage; This represents the actual losses due to evaporation, seepage, water intake, or other water volume losses included in the corresponding hydrological model. If a hydrological model does not include a certain loss item, that loss item will not be included in the calculation.

[0088] The water balance error limit shall be preferentially adopted from the allowable water balance residual specified in the hydrological model calibration report, acceptance report or business rules; if no such limit is specified, it shall be determined based on the combined measurement uncertainty or model calculation uncertainty of each water item.

[0089] Closed boundary state verification is used to determine whether the difference between the old and new values ​​of each model state variable at the dynamically recalculated closed boundary is still not greater than their respective convergence thresholds, and to determine whether the data sequences on both sides of the closed boundary are continuous.

[0090] Version consistency verification is used to determine whether the intermediate version, hydrological model version, business rule version, and river network topology version of the snapshot list record of the shadow application snapshot are consistent with the data lineage certificate of each application data partition, and to determine whether the application data partition that references the original application snapshot directly or indirectly references the changed intermediate version record.

[0091] If the closed boundary state verification fails, the current dynamic recalculation of the closed boundary is cancelled, and the incremental recalculation range is extended along the calculation time period. If the upstream and downstream continuity verification or water balance verification fails, the application data partitions of the hydrological nodes at both ends of the river segment that failed the verification within the corresponding influence time interval are added to the target failure set, and the incremental recalculation is re-executed from the nearest model state checkpoint that meets the conditions.

[0092] If the version consistency check fails but the application calculation result check passes, the scope of hydrological recalculation is not expanded. Instead, a new snapshot list and corresponding data lineage certificate for the shadow application snapshot are generated. If any check still fails after re-executing the corresponding process, the original application snapshot is kept as a valid snapshot, and the shadow application snapshot is marked as pending review.

[0093] After all hydrological consistency checks pass, the system atomically switches the valid snapshot pointer of the application snapshot layer from the original application snapshot to the shadow application snapshot.

[0094] Before the effective snapshot pointer is switched, business queries continue to read the original application snapshot; after the switch is complete, newly initiated business queries read the shadow application snapshot. Queries already bound to the original application snapshot continue to use the original application snapshot until the end of the current query, avoiding the reading of two different versions of application data in the same business query.

[0095] The original application snapshot will continue to be saved according to the snapshot retention rules, for comparison of results before and after revision, historical tracing, and rollback. Switching application snapshots will not delete historical hydrological observation events in the original evidence layer, nor will it delete expired intermediate version records in the intermediate version layer.

[0096] Taking periodically collected rainfall data as an example, rain gauge A should generate a rainfall measurement at the observation time, but fails to receive the measurement within the allowed delay period. The system generates an intermediate version record with a quality status of "missing measurement pending" based on the sequence of expected observation times.

[0097] Subsequently, the system receives the actual rainfall data corresponding to the observation time of rain gauge station A. Since the observation time of this data is earlier than the difference between the time boundary of the processed event and the allowable lateness time, the system identifies the data as a late event and determines the original missing intermediate version record as the intermediate version record to be changed.

[0098] After standardization and quality verification, late events are recorded as new intermediate versions. The system calls the corresponding river network topology version based on the observation time of the event, starting from the catchment node corresponding to rain gauge A, and forms a hydrological propagation cone based on the confluence response time, river segment propagation time, storage response time, model state memory time, and application data accumulation time window.

[0099] The system combines hydrological propagation cones with data lineage certificates, and only adds application data partitions that are located within the affected hydrological nodes and time intervals and that directly or indirectly reference the original missing intermediate version records to the target failure set.

[0100] Incremental recalculation begins at the most recent eligible model state checkpoint before the start time of the impact. When a river branch satisfies the model state convergence condition for a specified number of consecutive calculation periods, the upstream contribution data summary is consistent, and there are no unprocessed late events, a dynamic recalculation closure boundary is formed for that river branch. Application data partitions after the closure boundary continue to reference data from the original application snapshot, without performing full-basin and full-period recalculation.

[0101] Taking the verification of flow measurement results as an example, the manually verified results have the same hydrological spatial object identifier, observation element identifier and observation time as the original measurement results, but the source data versions or data content summaries of the two are different. The system will identify the manually verified results as a revision event.

[0102] The results of manual verification are incorporated into the original evidence layer as new hydrological observation events. After standardization and quality verification, new intermediate version records are generated. The intermediate version records corresponding to the original measurement results are identified as the modified intermediate version records, and a substitution relationship between the new intermediate version records and the modified intermediate version records is established.

[0103] The system calls the river network topology version that was valid at the time the flow measurement occurred. Even if the flow measurement section subsequently relocates or the river segment division is adjusted, the historical flow revision still determines the hydrological propagation cone according to the hydrological nodes and river segment relationships at the time the flow measurement occurred.

[0104] The system determines the target failure set based on the hydrological propagation cone and data lineage credentials, and begins incremental recalculation starting from the model state checkpoints that meet the criteria. The incremental recalculation results are written to the shadow application snapshot, and after passing the hydrological consistency check, they are switched to the valid application snapshot. The original data before the revision, old intermediate version records, and the original application snapshot are retained for traceability.

[0105] Allowable late arrival time is recorded in the data source configuration version; value range, rate of change limit, continuity error limit, and water balance error limit are recorded in the verification rule version; model state variables, model allowable error, model state memory duration, and minimum stable duration are recorded in the hydrological model version; river segment propagation time, river direction, and node connection relationship are recorded in the river network topology version; application data cumulative time window and business accuracy limit are recorded in the business rule version; storage response time is associated with the scheduling rule version and the storage node operating status.

[0106] Each time an intermediate version record, model state checkpoint, data lineage certificate, and application snapshot are generated, the actual parameter version used is recorded. When historical data is recalculated, the parameter version corresponding to that historical data and the current recalculation task is called, and the updated current parameters are not directly used.

[0107] This solution can be verified using historical hydrological processes. A historical hydrological process containing late or revised events is selected, and both full recalculation and incremental recalculation (as described in this embodiment) are performed. The number of application data partitions read, the number of hydrological nodes and calculation periods involved, the amount of data transferred across layers, and the model runtime are recorded for both methods. The application results from the two methods are then compared. When the difference between the results of the two methods does not exceed the business accuracy limit specified in the business rule version, the incremental recalculation result is confirmed to be consistent with the full recalculation result. Specific test results should be obtained based on actual project operation data; a reduction ratio should not be pre-set when test data is lacking.

[0108] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A three-layer integrated storage method for multi-source hydrological survey data, comprising the following steps: S1: Receive multi-source hydrological survey data, parse hydrological spatial object identifiers, observation element identifiers, observation times, data source identifiers, and source data versions, generate hydrological observation event identifiers, and bind the hydrological observation event identifiers to the river network topology versions corresponding to the observation times; S2: Store the original data corresponding to the hydrological observation event in the original evidence layer in an append-only manner, perform standardization processing and quality verification on the hydrological observation event, and generate intermediate version records stored in the intermediate version layer; S3: Based on intermediate version records whose quality status is provisionally valid or officially valid and not marked as invalid, generate application data partitions according to hydrological nodes and calculation periods, organize the application data partitions into application snapshots stored in the application snapshot layer, and generate data lineage certificates that record the intermediate version records they reference and model state checkpoints that record the corresponding set of hydrological node model state variables for each application data partition. S4: When a late event or revision event is identified, a new intermediate version record is generated, the changed intermediate version record is identified as being supplemented or replaced by the new intermediate version record, and the hydrological propagation cone composed of the affected hydrological nodes and their time intervals is determined based on the river network topology version, river direction and hydrological propagation time corresponding to the late event or revision event. S5: The application data partition whose hydrological node and calculation period are located within the hydrological propagation cone and whose data lineage certificate directly or indirectly references the changed intermediate version record is identified as the target failure set; S6: Start incremental recalculation from the nearest model state checkpoint that is before the start time of the impact of the target failure set, does not reference the intermediate version record affected by the change, and uses the same hydrological model version and river network topology version as this incremental recalculation. Determine the dynamic recalculation closure boundary based on whether the new model state variable set formed by the incremental recalculation and the old model state variable set in the original application snapshot continuously meet the convergence condition, and stop recalculating the application data partition after the dynamic recalculation closure boundary. S7: Write the application data partitions formed by incremental recalculation into the shadow application snapshot, and reference the application data partitions in the original application snapshot that do not belong to the target failure set to the shadow application snapshot. After the shadow application snapshot passes the closed boundary state and version relationship consistency verification, switch the effective application snapshot from the original application snapshot to the shadow application snapshot.

2. The three-layer integrated storage method for multi-source hydrological survey data according to claim 1, characterized in that, Multi-source hydrological survey data comes from automatic hydrological stations, manual measurement terminals, mobile flow measurement equipment, remote sensing data platforms, and historical compiled data; The hydrological spatial object identifier, observation element identifier, observation time, data source identifier, and source data version are concatenated according to a fixed field order. The concatenation result is then used to perform a summary operation on the data content summary of the multi-source hydrological survey data to obtain the hydrological observation event identifier.

3. The three-layer integrated storage method for multi-source hydrological survey data according to claim 2, characterized in that, When storing raw data in the raw evidence layer, structured measurements are written into the time series data area according to the hydrological spatial object identifier, observation element identifier, and observation time. Unstructured data is written into the object storage area, and an event index is established with the hydrological observation event identifier as the index key. The event index records the data partition where the structured measurement value is located, the object address and data content summary of the unstructured data, the data reception time, and the data entry batch; For new received data with the same hydrological spatial object identifier, observation element identifier, and observation time, the new hydrological observation event is written into the original evidence layer without covering the existing hydrological observation event.

4. The three-layer integrated storage method for multi-source hydrological survey data according to claim 3, characterized in that, Standardization processes include unit conversion, time reference conversion, spatial object encoding mapping, and missing data identifier conversion; Based on the data source category, observation element category, hydrological spatial object category, and acquisition method of hydrological observation events, quality verification items are determined from the applicable relationships of quality verification. Value range verification and time continuity verification are performed on structured measurements, change rate verification is performed on structured measurements with continuous sampling time sequence, upstream and downstream hydrological relationship verification is performed on structured measurements located at river network connectivity nodes, and equipment operation status verification is performed on structured measurements automatically acquired by monitoring equipment. If the currently executable quality verification project passes but the verification time window has not yet ended, an intermediate version record with a provisionally valid quality status is generated; when the verification time window ends and all applicable quality verification projects pass, an intermediate version record with a formally valid quality status is generated. The intermediate version record also records the verification rule version, equipment calibration version, river network topology version, and the identifier of the previous intermediate version, and marks the previous intermediate version record that is replaced by the new intermediate version record as invalid.

5. The three-layer integrated storage method for multi-source hydrological survey data according to claim 4, characterized in that, For each combination of data source identifier, hydrological spatial object identifier, and observation element identifier, maintain the time boundary of processed events and the allowable delay duration. The time boundary of processed events is the maximum observation time of the intermediate version record in the corresponding combination when the quality status reaches the provisional or formal validity. The allowable delay duration is determined according to the acquisition cycle, reporting cycle, and retransmission deadline of the corresponding acquisition source. When the observation time of newly received hydrological survey data is earlier than the difference between the time boundary of the processed event and the allowable lateness, it is identified as a late event, and the missing intermediate version record that has the same hydrological spatial object identifier, observation element identifier and observation time as the late event is identified as the intermediate version record to be changed. When newly received hydrological survey data and existing hydrological observation events have the same hydrological spatial object identifier, observation element identifier, and observation time, and the source data versions or data content summaries of the two are different, the newly received hydrological survey data will be identified as a revision event, and the intermediate version record corresponding to the existing hydrological observation event will be identified as the intermediate version record subject to change.

6. The three-layer integrated storage method for multi-source hydrological survey data according to claim 5, characterized in that, The river network topology version records hydrological nodes, river segments, the connection relationships between nodes and river segments, river direction, river segment propagation time, and the effective time interval of the topology version. Starting from the hydrological node where the late event or revised event is located, traverse the downstream hydrological nodes along the river direction, and determine the sum of the observation time of the late event or revised event and the cumulative river segment propagation time to the current downstream hydrological node as the start time of the impact of the current downstream hydrological node. The impact termination time of the current downstream hydrological node is determined based on the basin confluence response duration, model state memory duration, cumulative application data time window length, and storage response duration of storage nodes in the propagation path. The impact time interval of the current downstream hydrological node is formed by the impact start time and impact termination time.

7. The three-layer integrated storage method for multi-source hydrological survey data according to claim 6, characterized in that, The data lineage certificate also records the upstream application data partition identifier, hydrological model version, business rule version, river network topology version, and output result summary referenced by the application data partition; When a data lineage certificate record is modified by an intermediate version record, the corresponding application data partition is determined to directly reference the modified intermediate version record. When the data lineage certificate records the upstream application data partition, and the data lineage certificate of the upstream application data partition is traced step by step to obtain the changed intermediate version record, it is determined that the corresponding application data partition indirectly references the changed intermediate version record. The application data partition that directly or indirectly references the changed intermediate version record and belongs to the affected hydrological nodes in the hydrological propagation cone, is added to the target failure set.

8. The three-layer integrated storage method for multi-source hydrological survey data according to claim 7, characterized in that, The model status checkpoint also records the upstream contribution data summary, hydrological model version, river network topology version, and application snapshot version; The upstream contribution data summary is generated based on the intermediate version identifier referenced by the application data partition, the upstream application data partition identifier, the output result summary of the upstream application data partition, the hydrological model version, and the river network topology version; If there are no unreferenced intermediate version records before the start time of the impact and the model status checkpoints that use the same hydrological model version and river network topology version as this incremental recalculation, the starting state of the hydrological calculation cycle containing the start time of the impact shall be used as the starting point of the incremental recalculation. Incremental recalculation is performed according to the order of calculation time periods and river direction.

9. A three-layer integrated storage method for multi-source hydrological survey data according to claim 8, characterized in that, The convergence threshold for each model state variable is determined based on the allowable state deviation specified in the corresponding hydrological model version and the allowable deviation specified in the business rules version; The number of consecutive convergence calculation periods is determined based on the minimum stable duration of the hydrological model version and the model calculation step size, and the number of consecutive convergence calculation periods is not less than two. When the difference between the old and new values ​​of each model state variable is not greater than its corresponding convergence threshold, the new and old upstream contribution data summaries are consistent, and there are no unprocessed late events or revision events in each calculation period and its upstream propagation path, the last calculation period is determined as the dynamic recalculation closure boundary. When the river network has branches, the dynamic recalculation closure boundary is determined for each branch, and the incremental recalculation range is stopped from being expanded for branches that have formed dynamic recalculation closure boundaries.

10. A three-layer integrated storage method for multi-source hydrological survey data according to claim 9, characterized in that, Shadow application snapshots record application data partitions formed by incremental recalculation and application data partitions referenced from the original application snapshots through a snapshot list; The shadow application snapshots are subjected to upstream and downstream continuity verification, water balance verification, closed boundary state verification, and version consistency verification. Among them, the upstream and downstream continuity verification determines whether the difference between the output flow of the upstream hydrological node after the river segment propagation time alignment and the input flow component of the downstream hydrological node originating from the upstream hydrological node exceeds the continuity error limit. The water balance verification determines whether the water balance residual formed by the inflow, interval inflow, outflow, water storage change and water loss exceeds the water balance error limit. The closed boundary state verification determines whether the difference between the old and new values ​​of the model state variables at the dynamically recalculated closed boundary exceeds the corresponding convergence threshold. The version consistency verification determines whether the intermediate version, hydrological model version and river network topology version recorded in the snapshot list are consistent with the data lineage certificate of each application data partition. When the closed boundary state verification fails, the incremental recalculation range is extended along the calculation period direction. When the upstream and downstream continuity verification or water balance verification fails, the application data partitions of the hydrological nodes at both ends of the river section that failed the verification are added to the target failure set within the corresponding impact time interval. When the version consistency verification fails, the snapshot list and data lineage certificate are regenerated. After all checks pass, the valid snapshot pointer is switched atomically from the original application snapshot to the shadow application snapshot.

Images